Changes to the setting that hides the built-in artifact for MCP

What

Claude Code has a built-in artifact tool, and it can hide it in favour of an artifact tool that comes through MCP (Model Context Protocol, a way to plug outside tools into Claude Code). The setting for this is builtInArtifactHiddenForMcp. Two things around it changed:

• Resetting it does more: before, the reset only cleared builtInArtifactHiddenForMcp and live. It now also sets autoReact.optIn and autoReact.wakeEnabledMemo to false, and calls handBackTakenOverSlug for each entry in wakes.takenFrom, which hands back any wakes that had been taken over.
• The failed-watch message is shorter: when a live subscription to an artifact fails to start, Claude is told about it. When builtInArtifactHiddenForMcp is set, the message now leaves out its suggested remedy and only says 'do not claim to be watching it'. Before, the remedy was always included.

Why

Turning the setting off now leaves no auto-react or wake state behind. Claude also no longer gets a suggested fix that may not apply when the built-in artifact is hidden, but it is still told not to claim it is watching an artifact it is not.